an analysis of the exponential family models to predict yield loss of safflower (carthamus tinctorius l.) challenged with water stress and redroot pigweed (amaranthus retroflexus l.)

the performance of different yield loss models from an exponential family was evaluated in safflower-redroot pigweed systems in two field experiments conducted during 2007 and 2008 growing seasons at the research field of agricultural college of shiraz university, iran. the yield loss of safflower was recorded as relative yield loss in experimental plots laid out in split plot design with three replicates. three different irrigation treatments were allocated to the main plots and consisted of full irrigation or 100% field capacity (fc), 75% fc, and 50% fc, while five weed densities (0, 3, 6, 9, and 12 weeds m-2) were assigned to the sub-plots the logistic and gompertz models and a user defined power-exponential model were fitted to the data to relate crop yield loss to the weed densities under different water stress conditions. the power-exponential model was chosen as the best fit to the data with statistically acceptable model diagnostics. logistic and gompertz models showed good fit to the observed data, but underestimated the yield loss under three levels of irrigation. model performance in all cases was influenced by water stress as models generally showed greater constant and systematic biases under severe water stress (50% fc). model parameters were used to explain the impact of water stress in crop/weed system. the exponential family models globally performed better over common empirical models such as spitters, kropff and lotz and cousens models.